Suicide Risk Detection Model

This project builds a machine learning model using an LSTM (Long Short-Term Memory) neural network for detecting suicidal tendencies in text data. The model is trained on a labeled dataset of text samples and uses pre-trained GloVe embeddings for word representations.

Table of Contents

Table of Contents

• Overview
• Dataset
• Model Architecture
• Installation & Dependencies
• IPyWidgets for Interaction
• Comparison of Different Models

Overview

The goal of this project is to detect suicide-related tendencies in textual data using a deep learning model. The model is trained on a dataset that contains user-generated text labeled as either suicidal or non-suicidal.

The architecture utilizes GloVe embeddings for word representations and an LSTM layer for sequential text processing.

Dataset

The dataset Suicide Detection Dataset(Suicide_Detection.csv) consists of user-generated text labeled as either suicide or non-suicide. For this project, text cleaning and preprocessing are essential steps.

or You can dowload the dataset from this drive link including the Glove Embedding(6B,300-Dimensional Vectors):

-Drive Link: Suicide Detection Dataset

Features

• Text: User-generated posts or comments.
• Class: Labels indicating whether the text has suicidal tendencies (suicide) or not (non-suicide).
• Important: You need to download or prepare a dataset and save it as Suicide_Detection.csv before running the code.

Model Architecture

The model is a sequential deep learning model with the following layers:

• Embedding Layer: Uses pre-trained GloVe embeddings (6B, 300-dimensional vectors) to transform words into dense vectors.
• LSTM Layer: An LSTM with 64 units to capture temporal dependencies in the text.
• GlobalMaxPooling1D Layer: To capture the most important features across the sequence.
• Dense Layer: Fully connected layers for classification.
• Dropout Layers: To prevent overfitting.
• Output Layer: A single neuron with a sigmoid activation function for binary classification (suicidal or non-suicidal).

Installation & Dependencies

Clone the repository:

git clone https://github.com/TajKab/Suicide_Risk_Analyzer.git
cd suicide-risk-detection

Install the necessary dependencies:

pip install -r requirements.txt

Dependencies:

• keras
• tensorflow
• numpy
• pandas
• matplotlib
• sklearn
• pickle

Interactive Output

IPyWidgets for Interaction:

The code includes IPyWidgets for interactive text analysis. The widget allows users to input a sentence and receive an immediate analysis indicating whether the sentence suggests suicidal tendencies.

Comparison of Different Models

Comparison of Models In this project, we explored both deep learning and traditional machine learning models for suicide detection using text data.

• LSTM-based NLP Model: Utilizes word embeddings (GloVe) and an LSTM layer for text classification. Achieved high accuracy and recall for detecting suicidal tendencies.
• Traditional Machine Learning Models: TF-IDF Vectorization: Text data was also transformed using TF-IDF, a common technique for converting text into numerical features. Three traditional machine learning models were used: Logistic Regression, Decision Tree, and Random Forest*. The following table presents the performance metrics (Accuracy, F1 Score, Precision, and Recall) for all models:

The differences in outputs for traditional models stem primarily from how they handle the input features (TF-IDF vs. padded sequences)